Class "msfit"

Description

Stores the output of Bayesian variable selection, as produced by function modelSelection. The class extends a list, so all usual methods for lists also work for msfit objects, e.g. accessing elements, retrieving names etc.

Some additional methods are provided for printing information on screen, computing posterior probabilities or sampling from the posterior of regression coefficients, as indicated below.

Objects from the Class

Typically objects are automatically created by a call to modelSelection. Alternatively, objects can be created by calls of the form new("msfit",x) where x is a list with the adequate elements (see slots).

Slots

The class extends a list with elements:

postSample

matrix with posterior samples for the model indicator. postSample[i,j]==1 indicates that variable j was included in the model in the MCMC iteration i

postOther

postOther returns posterior samples for parameters other than the model indicator, i.e. basically hyper-parameters. If hyper-parameters were fixed in the model specification, postOther will be empty.

margpp

Marginal posterior probability for inclusion of each covariate. This is computed by averaging marginal post prob for inclusion in each Gibbs iteration, which is much more accurate than simply taking colMeans(postSample)

.

postMode

Model with highest posterior probability amongst all those visited

postModeProb

Unnormalized posterior prob of posterior mode (log scale)

postProb

Unnormalized posterior prob of each visited model (log scale)

coef

Estimated coefficients (via posterior mode) for highest posterior probability model

family

Residual distribution, i.e. argument family when calling modelSelection

p

Number of variables

Methods

show

signature(object = "msfit"): Displays general information about the object.

postProb

signature(object = "msfit"): Extracts posterior model probabilities.

rnlp

signature(object = "msfit"): Obtain posterior samples for regression coefficients.

Author(s)

David Rossell

References

Johnson VE, Rossell D. Non-Local Prior Densities for Default Bayesian Hypothesis Tests. Journal of the Royal Statistical Society B, 2010, 72, 143-170

Johnson VE, Rossell D. Bayesian model selection in high-dimensional settings. Journal of the American Statistical Association, 107, 498:649-660.

See Also

See also modelSelection and rnlp.

Examples

1
showClass("msfit")